1980s

The 1980s were a formative decade for the research program that made intuitive judgment visible as a source of systematic error rather than merely occasional irrationality. During this period, the Linda problem became a celebrated and controversial demonstration of the conjunction fallacy, while the broader framework of heuristics and biases connected experiments on representativeness, availability, substitution, probability, and decision-making to a changing account of how the mind works.

The Decade and the Research Program

The work associated with Daniel Kahneman and Amos Tversky had been developing for several years before the 1980s. Their collaboration examined how people assign probabilities, forecast the future, assess hypotheses, and estimate frequencies. In the fifth year of their collaboration, they presented their principal findings in the journal Science in an article titled “Judgment Under Uncertainty: Heuristics and Biases.” The article described intuitive shortcuts and treated roughly twenty recurring biases as manifestations and demonstrations of those shortcuts.

The significance of this program lay in what it challenged. Social scientists in the 1970s had broadly accepted that people were generally rational and that departures from rationality were chiefly explained by emotions such as fear, affection, or hatred. The heuristics-and-biases research offered a different diagnosis: normal people could make systematic, predictable errors because of the design of the machinery of cognition itself. The work attracted more attention than its authors expected and later became one of the most highly cited works in social science, with more than three hundred scholarly articles referring to it in 2010. Its concepts were subsequently used in medical diagnosis, legal judgment, intelligence analysis, philosophy, finance, statistics, and military strategy.

Intuition and Question Substitution

A central idea developed in this period was question substitution. People normally possess intuitive feelings and opinions about nearly everything, including people they barely know, strangers they cannot explain trusting or distrusting, and enterprises they feel certain will succeed without analyzing them. When a difficult target question does not receive a satisfactory answer quickly, System 1 searches for a related question that is easier and answers that instead. The intended assessment is the target question; the easier substitute is the heuristic question.

This mechanism explains why intuitive judgments can feel complete even when the underlying problem has not been properly understood. The mind avoids being stumped by producing an answer to a nearby question, and the resulting opinion may be coherent, vivid, and confidently held without being logically responsive to the original task. The same research tradition organized such errors around three major heuristics: representativeness, used when judging whether an object or event belongs to a class or process; availability, used when assessing the frequency of a class or the plausibility of a development; and adjustment from an anchor, used in numerical prediction when a relevant value is available. These shortcuts are economical and often effective, but they can also produce systematic and predictable mistakes.

The Linda Problem and the Conjunction Fallacy

The best-known experiment of the period presented Linda as a thirty-one-year-old single woman who was outspoken and bright, had majored in philosophy, had been concerned with discrimination and social justice, and had participated in antinuclear demonstrations. Audiences in the 1980s often laughed because the description seemed to identify her immediately with the politically engaged students associated with the University of California at Berkeley. The portrait made some occupations and affiliations seem representative of Linda and others seem sharply inconsistent with her character.

Participants were asked to rank possible descriptions, including a teacher in elementary school, a bookstore worker who took yoga classes, an active feminist, a psychiatric social worker, a member of the League of Women Voters, a bank teller, an insurance salesperson, and a bank teller active in the feminist movement. The crucial comparison was between “Linda is a bank teller” and “Linda is a bank teller and is active in the feminist movement.” The second description formed a more coherent story and fit the stereotype generated by the portrait, so it was judged more representative even though a conjunction cannot be more probable than either of its component events.

The error survived efforts to make the logical issue unmistakable. About 85% to 90% of undergraduates at several major universities selected the conjunction, and in one sample 89% of undergraduates violated the probability rule. Even doctoral students in the decision-science program of the Stanford Graduate School of Business, all of whom had taken advanced courses in probability, statistics, and decision theory, showed the same tendency: 85% ranked the feminist bank teller as more likely than the bank teller. The experimenters’ attempts to remove the error became increasingly desperate, but the intuitive story remained more persuasive than the elementary rule.

The resulting term, conjunction fallacy, names the judgment that a conjunction of two events is more probable than one of those events considered alone. The fallacy was not simply a failure to understand probability; it was an attractive judgment produced by the coherence of the narrative. The effect could remain compelling even after participants recognized the logical answer. Stephen Jay Gould described knowing the correct response while still hearing an insistent inner voice protesting that Linda could not merely be a bank teller. The episode became an emblem of the tension between the story-building force of System 1 and the corrective demands of System 2.

From Formal Probability to Human Evaluation

The decade’s research also sharpened the contrast between intuitive judgment and Bayesian discipline. Bayes’s rule requires prior beliefs, or base rates, to be combined with the diagnosticity of evidence. If 3% of graduate students are enrolled in computer science and a description of Tom W is four times more likely for a computer-science student than for students in other fields, the resulting probability that Tom W is a computer scientist is 11%; if the base rate is 80%, the result is 94.1%. The examples show why evidence about an individual case does not erase information about the population from which the case comes, and why people often exaggerate how diagnostic a vivid description is.

This concern with formal constraints extended into the study of risky choice. Prospect theory treated gains and losses, rather than total wealth, as the effective carriers of subjective value. Evaluation is relative to a reference point, usually the status quo but sometimes an expected or entitled outcome. The theory also described diminishing sensitivity: the subjective difference between $900 and $1,000 is smaller than the difference between $100 and $200. Its third central principle was loss aversion, the tendency for losses to loom larger than corresponding gains. The resulting value function is S-shaped, with diminishing sensitivity on both sides of the reference point and a sharper response to losses than to gains.

The fourfold pattern extended this analysis to the weighting of uncertain outcomes. When people evaluate a complex object or an uncertain prospect, System 1 assigns different weights to different characteristics or possible outcomes, whether or not the person is aware of doing so. Those weights correlate with probabilities but do not simply reproduce them. A 50% chance of winning a million dollars is more attractive than a 1% chance of winning the same amount, yet the psychological treatment of probabilities is not a straightforward mathematical calculation.

Experience, Memory, and Retrospective Time

Research reported in the book also distinguished momentary experience from retrospective evaluation. In the colonoscopy study designed by Daniel Kahneman and Don Redelmeier, 154 patients reported their pain every sixty seconds on a scale from zero to ten. One illustrated procedure lasted eight minutes and another twenty-four minutes; the shortest procedure in the full sample lasted four minutes and the longest sixty-nine. If suffering were calculated as the area under the curve, the longer procedure would clearly have been worse.

Patients’ retrospective judgments followed a different logic. Their global ratings were well predicted by the average of the worst moment and the ending, a pattern called the peak-end rule. Duration neglect meant that the length of the procedure had no effect on ratings of total pain. Both illustrated patients reached a peak rating of eight, but the final rating was seven for the shorter procedure and one for the longer procedure. Their peak-end averages were therefore 7.5 and 4.5, so the shorter procedure left the more unpleasant memory despite being less painful in total.

The cold-hand experiment made the conflict between experience and memory a basis for choice. Participants held a hand in painfully cold water while continuously recording their momentary pain. The short trial lasted sixty seconds at 14 degrees Celsius. The long trial began identically but continued for another thirty seconds while slightly warmer water reduced the pain by roughly one degree. Although the long episode was objectively worse because it contained all of the short episode plus additional pain, 80% of participants who experienced a decline in pain at the end chose to repeat the long trial. Their decision was governed by the memory of the experience, not by its duration-weighted total. The decade’s work thus treated time not only as elapsed duration but also as a pattern reconstructed by the remembering self.